For best outputs and usage:

This model descends from Pony Diffusion XL, and in big part it's like it, but in turbo mode.

Sampler: DPM++ SDE.

Steps: 6 ~ 8, cfg 2, CLIP skip 2.

Pos prompt: always start your pos with "score_9, score_8_up, score_7_up," then your text. Optionally, you can use the LoRa "2.5DRealistic" with a weight of 0.5 ~ 0.6 (included in the semi-realistic example runs) to increase the output's realism, or use the LoRa "2.5DCartoonV2" with a similar weight (included in the non-anime cartoon example runs) to give it a cartoon like look. By default without the LoRa's, the model tends to create anime like outputs.

Neg prompt: "score_4, score_5, score_6, FastNegativeV2, BadDream, BeyondSDXLv3", being "score_4, score_5, score_6" mandatory, and the embeddings "FastNegativeV2", "BadDream" and "BeyondSDXLv3" optional but highly recommended to increase overall quality (included in all my example runs).